#952: gcc should be passed the -fwrapv flag
----------------------+-----------------------------------------------------
Reporter: simonmar | Owner:
Type: bug | Status: reopened
Priority: normal | Milestone: 6.6.1
Component: Compiler | Version: 6.6
Severity: normal | Resolution:
Keywords: | Difficulty: Unknown
Testcase: | Architecture: Multiple
Os: Multiple |
----------------------+-----------------------------------------------------
Changes (by simonmar):
* resolution: fixed =>
* summary: Floating point exception building HEAD with 6.6 => gcc should
be passed the -fwrapv flag
* status: closed => reopened
Comment:
re-opening. After giving this some more thought, I realised that the
Haskell report's meaning of "undefined" is different from the C language
specification's meaning of "undefined", and that in fact `ghc -fvia-C` is
behaving incorrectly here.
Overflow of `Int` should return either `_|_` or an implementation-defined
value, according to the Haskell report. In the code affected by this bug,
we were returning an implementation-defined value (2's complement
wrapping) but a subsequent test on the value was returning an inconsistent
answer; this is allowed by the C spec, but not by Haskell.
The upshot is that I believe we should invoke gcc with `-fwrapv` where it
is supported (gcc 3.4+). This is more than required: it forces 2's
complement wrapping behviour, but prevents the illegal optimisations that
give rise to this bug.
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/952>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler_______________________________________________
Glasgow-haskell-bugs mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
